credo-release

Pass

Audited by Gen Agent Trust Hub on Sep 10, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [SAFE]: The primary component is a Python script designed to perform static analysis on LaTeX documents to find common errors such as TODO markers, merge conflicts, and doubled words.\n- [COMMAND_EXECUTION]: The regression test file scripts/test_check_tex.py utilizes subprocess.run to execute the main script check_tex.py. This is used exclusively for internal validation of the linter's logic and does not incorporate untrusted input into the shell command.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted text data from user-provided LaTeX files and build logs. While this creates a surface for indirect prompt injection, the skill's capabilities are highly restricted.\n
  • Ingestion points: Content is read from local file paths specified by the user or found via LaTeX \\input and \\include commands in scripts/check_tex.py.\n
  • Boundary markers: The agent is instructed to report findings verbatim from the script's output, which focuses on specific mechanical errors.\n
  • Capability inventory: The analysis script is restricted to reading files and printing to standard output; it cannot write files, access the network, or perform privilege escalation.\n
  • Sanitization: The script includes logic to strip LaTeX comments (%) which prevents commented-out text from triggering false positives in the linting process.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 10, 2026, 11:52 AM
Security Audit — agent-trust-hub — credo-release